Causal Mechanisms in the Social Sciences

SO36CH03-Hedstrom ARI 7 June 2010 22:40

Causal Mechanismsin the Social SciencesPeter Hedstrom1 and Petri Ylikoski21Nuffield College, University of Oxford, Oxford, OX1 1NF United Kingdom;email: [email protected] of History and Philosophy, 33014 University of Tampere, Finland;email: [email protected]

Annu. Rev. Sociol. 2010. 36:49–67

First published online as a Review in Advance onApril 5, 2010

The Annual Review of Sociology is online atsoc.annualreviews.org

This article’s doi:10.1146/annurev.soc.012809.102632

Copyright c© 2010 by Annual Reviews.All rights reserved

0360-0572/10/0811-0049$20.00

Key Words

explanation, causal inference, realism, agent-based simulation,individualism, middle-range theory

Abstract

During the past decade, social mechanisms and mechanism-based ex-planations have received considerable attention in the social sciences aswell as in the philosophy of science. This article critically reviews themost important philosophical and social science contributions to themechanism approach. The first part discusses the idea of mechanism-based explanation from the point of view of philosophy of science andrelates it to causation and to the covering-law account of explanation.The second part focuses on how the idea of mechanisms has been usedin the social sciences. The final part discusses recent developments inanalytical sociology, covering the nature of sociological explananda, therole of theory of action in mechanism-based explanations, Merton’s ideaof middle-range theory, and the role of agent-based simulations in thedevelopment of mechanism-based explanations.

49

Ann

u. R

ev. S

ocio

l. 20

10.3

6:49

-67.

Dow

nloa

ded

from

ww

w.a

nnua

lrev

iew

s.or

gby

Uni

vers

ity o

f C

alif

orni

a -

Sant

a C

ruz

on 0

9/16

/12.

For

per

sona

l use

onl

y.


INTRODUCTION

During the past decade, social mechanisms andmechanism-based explanations have receivedconsiderable attention in the social sciencesas well as in the philosophy of science. Somewriters have described this as a mechanismmovement that is sweeping the social sciences(Norkus 2005). Although that is not an entirelycorrect description, the literature on mecha-nisms has grown rapidly, and it touches uponmany fundamental questions related to the aimof the social sciences.

The idea that science aims to providemechanism-based explanations of phenomenahas a long history (Bechtel 2006), but only inrecent decades has this idea been an objectof more systematic study. In the philosophyof science, mechanism-based explanations havebeen discussed mainly in the context of bio-logical sciences (Bechtel & Richardson 1993;Bechtel 2006, 2008; Craver 2007; Darden 2006;Glennan 2002; Thagard 1999; Wimsatt 2007),whereas in the social sciences the idea hasbeen mostly discussed by social scientists them-selves (Abbott 2007; Elster 1989, 2007; Gross2009; Hedstrom 2005; Hedstrom & Swedberg1998a; Manicas 2006; Mayntz 2004; Morgan &Winship 2007; Schmidt 2006; Tilly 2001;Wikstrom 2006). Although the basic ideas arequite similar, the discussions within philosophyand social science have proceeded mostly inde-pendently of each other. In both contexts, thedevelopment of the idea of a mechanism-basedexplanation has been partly motivated by theshortcomings of the once hegemonic covering-law account of explanation (Hempel 1965) andof purely statistical explanations. The basic ideaof a mechanism-based explanation is quite sim-ple: At its core, it implies that proper explana-tions should detail the cogs and wheels of thecausal process through which the outcome tobe explained was brought about.

In this article, we critically review the mostimportant philosophical and social sciencecontributions to the mechanism approach.The article is organized as follows. In thenext section, we discuss the ideas of causal

mechanisms and mechanism-based explana-tions from a philosophy-of-science point ofview. Thereafter, we review some of the uses ofmechanism-based ideas in the social sciences.Finally, we focus on analytical sociologyand show how ideas about mechanism-basedexplanations fit into a broader view of the aimsand standards of sociology.

THE IDEA OFMECHANISM-BASEDEXPLANATIONS

Because the entities and processes studied bydifferent sciences are quite heterogeneous, itis difficult to propose a mechanism definitionthat would both be informative and cover all ex-amples of mechanisms. Some disciplines, suchas cell biology (Bechtel 2006) and the neuro-sciences (Craver 2007), study highly integratedsystems, whereas others, such as evolutionarybiology and the social sciences, study more dis-persed phenomena. For this reason, a charac-terization of a mechanism that applies to onefield might not be informative when applied toanother (Skipper & Millstein 2005, Kuorikoski2009).

However, some general ideas are sharedby most accepted mechanism definitions (seeTable 1). First, a mechanism is identified bythe kind of effect or phenomenon it produces.A mechanism is always a mechanism for some-thing (Darden 2006, p. 273). The characteri-zation of the effect often requires some care.For example, a standard roulette wheel doesnot have different mechanisms for distributingthe ball to pockets 16 and 17. Rather, the samemechanism produces all 37 outcomes.

Second, a mechanism is an irreduciblycausal notion. It refers to the entities of acausal process that produces the effect ofinterest. Some mechanism definitions makeunnecessarily strong assumptions, however.For example, when Mahoney (2001, p. 580)characterizes a causal mechanism as an unob-servable entity that is sufficient to produce theoutcome of interest, he makes two unnecessarycommitments. First, there is nothing in the

50 Hedstrom · Ylikoski

Ann

u. R

ev. S

ocio

l. 20

10.3

6:49

-67.

Dow

nloa

ded

from

ww

w.a

nnua

lrev

iew

s.or

gby

Uni

vers

ity o

f C

alif

orni

a -

Sant

a C

ruz

on 0

9/16

/12.

For

per

sona

l use

onl

y.


Table 1 Alternative mechanism definitions

Author Definition ReferencesBechtel &Abrahamsen

A mechanism is a structure performing a function by virtue of its componentparts and component operations and their organization. The orchestratedfunctioning of the mechanism is responsible for one or more phenomena.

Bechtel & Abrahamsen 2005;Bechtel 2006, 2008

Bunge A mechanism is a process in a concrete system that is capable of bringingabout or preventing some change in the system.

Bunge 1997, 2004

Glennan A mechanism for a behavior is a complex system that produces that behaviorby the interaction of several parts, where the interactions between parts canbe characterized by direct, invariant, change-relating generalizations.

Glennan 2002

Machamer, Darden,and Craver

Mechanisms are entities and activities organized such that they produceregular changes from start to finish.

Machamer et al. 2000,Darden 2006, Craver 2007

Elster I A mechanism explains by opening up the black box and showing the cogs andwheels of the internal machinery. A mechanism provides a continuous andcontiguous chain of causal or intentional links between the explanans andthe explanandum.

Elster 1989

Elster II Mechanisms are frequently occurring and easily recognizable causal patternsthat are triggered under generally unknown conditions.

Elster 1999

Hedstrom Mechanisms consist of entities (with their properties) and the activities thatthese entities engage in, either by themselves or in concert with otherentities. These activities bring about change, and the type of changebrought about depends on the properties of the entities and how the entitiesare organized spatially and temporally.

Hedstrom 2005

Little A causal mechanism is a series of events governed by law-like regularities thatlead from the explanans to the explanandum.

Little 1991

Woodward A model of a mechanism (a) describes an organized or structured set of partsor components, where (b) the behavior of each component is described by ageneralization that is invariant under interventions, and where (c) thegeneralizations governing each component are also independentlychangeable, and where (d ) the representation allows us to see how, by virtueof (a), (b), and (c), the overall output of the mechanism will vary undermanipulation of the input to each component and changes in thecomponents themselves.

Woodward 2002

notion of a mechanism that would implythat it is by definition unobservable. Most ofthe mechanisms constituting an automobile’sengine, for example, are quite visible whenone opens up the hood (Mayntz 2004). Ofcourse, when one appeals to mechanisms tomake sense of statistical associations, one isreferring to things that are not visible in thedata, but this is different from them beingunobservable in principle. Similarly, to requirethat the mechanism is sufficient for the effectis an all too strong requirement: A mechanismcan involve irreducibly stochastic elements andthus affect only the probability of a given effect.

Third, the mechanism has a structure.When a mechanism-based explanation opensthe black box, it discloses this structure. It turnsthe black box into a transparent box and makesvisible how the participating entities and theirproperties, activities, and relations produce theeffect of interest. For this reason, the suggestion(e.g., Opp 2005) that a mechanism is just an in-tervening variable misses an important point.The focus on mechanisms breaks up the origi-nal explanation-seeking why question into a se-ries of smaller questions about the causal pro-cess: What are the participating entities, andwhat are their relevant properties? How are the

www.annualreviews.org • Causal Mechanisms 51

Ann

u. R

ev. S

ocio

l. 20

10.3

6:49

-67.

Dow

nloa

ded

from

ww

w.a

nnua

lrev

iew

s.or

gby

Uni

vers

ity o

f C

alif

orni

a -

Sant

a C

ruz

on 0

9/16

/12.

For

per

sona

l use

onl

y.


interactions of these entities organized (bothspatially and temporally)? What factors couldprevent or modify the outcome? And so on.

Fourth, mechanisms form a hierarchy.While a mechanism at one level presupposesor takes for granted the existence of certain en-tities with characteristic properties and activ-ities, it is expected that there are lower-levelmechanisms that explain them. (For a discus-sion of different notions of mechanism-relatedlevels, see Craver 2007, chapter 5.) It is an in-herent feature of the mechanism view that theentities and mechanisms of various sciences areultimately related to each other. Although theexplanatory entities and mechanisms employedby one science always bottom out somewhere(Darden 2006, p. 26) and are therefore takenas fundamental, their fundamental status is rel-ative because they are mechanistically explain-able by other fields of science. Although it issometimes (Kincaid 1996, Norkus 2005, Opp2005) suggested that the idea of mechanism-based explanation leads to an infinite regress,this is not the case. First, for a mechanism tobe explanatory it is not required that the en-tities, properties, and activities that it appealsto are themselves explained. The only require-ment is that such entities, properties, and ac-tivities really exist; their explanation is a sep-arate question. Second, the mechanism-basedaccount is fully compatible with the notion thatthere are no mechanism-based explanations forfundamental (physical) processes.

There is an ambiguity in the use of the no-tion of a mechanism. Sometimes it is used to re-fer to a causal process that produces the effectof interest and sometimes to a representationof the crucial elements of such a process. Thisshould not be a cause of concern, however, be-cause the latter presupposes the former. Whenone makes a claim that a certain mechanism ex-plains some real world events, one commits tothe existence of the entities, properties, activ-ities, and relations that the description of themechanism refers to.

In the following, we follow a long philosoph-ical tradition and assume that explanations areanswers to questions (Hempel 1965, Salmon

1998, Woodward 2003). The why or howquestion one is addressing determines what therepresentation of the mechanism should in-clude in order to be explanatory. Only by know-ing the nature of the explanatory task at handcan one determine which details of a mecha-nism are relevant to include and the appropriatedegree of abstraction (Ylikoski 2010).

Roughly, mechanism-based explanationshave two kinds of explananda. First, they mightaddress particular empirical facts. In such cases,the description of the mechanism is often amodified adaptation and combination of moregeneral mechanism schemes. Second, theymight address stylized facts. Although the ex-planation of particular empirical facts is theultimate goal of mechanism-based theory de-velopment, most of the time theorists are ad-dressing highly stylized theoretical explanandathat do not necessarily have close resemblanceto any particular empirical fact. Consider forexample the small-world problem of Milgram(1967). Watts & Strogatz (1998) identifiedsome salient and abstract features of the small-world problem—average path length and localclustering—and developed a model that couldexplain them. The explanation of simplifiedand heavily idealized facts such as those identi-fied by Watts & Strogatz results in mechanismschemes with wide application domains. By es-tablishing how stylized facts can in principle beexplained, theorists contribute to the toolbox ofsemi-general mechanisms.

Following Darden (2006, p. 281), a mech-anism scheme can be defined as “a truncatedabstract description of a mechanism that canbe filled with more specific descriptions ofcomponent entities and activities.” In itself, amechanism scheme provides a how-possibleexplanation; it tells us how the effect could inprinciple be produced. As it is possible thatsimilar effects can be produced by a numberof different (known or unknown) mechanisms,a crucial element in any mechanism-based ex-planation of empirical facts is the collection ofempirical evidence about the assumed entities,activities, relations, etc. The empirical evidenceturns a possible mechanism into a plausible


Ann

u. R

ev. S

ocio

l. 20

10.3

6:49

-67.

Dow

nloa

ded

from

ww

w.a

nnua

lrev

iew

s.or

gby

Uni

vers

ity o

f C

alif

orni

a -

Sant

a C

ruz

on 0

9/16

/12.

For

per

sona

l use

onl

y.


mechanism and may eventually lead to theidentification of the actual mechanism. By pre-senting evidence in support of the assumptionsof one mechanism and showing the absenceof evidence for the assumptions of competingmechanisms, we increase the plausibility ofthe explanatory hypothesis. What separatesproper mechanism-based explanations frommere mechanism-based storytelling is this kindof rigorous checking of the assumptions uponwhich the mechanism schemes rest.

A mechanism-based explanation describesthe causal process selectively. It does not aimat an exhaustive account of all details but seeksto capture the crucial elements of the process byabstracting away the irrelevant details. The rel-evance of entities, their properties, and their in-teractions is determined by their ability to makea relevant difference to the outcome of interest.If the presence of an entity or of changes inits properties or activities truly does not makeany difference to the effect to be explained,it can be ignored. This counterfactual crite-rion of relevance implies that mechanism-basedexplanations involve counterfactual reasoningabout possible changes and their consequences(Ylikoski 2010).

Mechanisms and Causality

The metaphysics of causation is still hotlydebated among philosophers, so it is anadvantage that the mechanism-based accountof explanation is not wedded to a specifictheory of causation. However, the mechanismperspective sets some important constraints foran acceptable theory of causation. For example,the idea of productive causal activities associ-ated with the mechanism perspective implies acommitment to the locality of causal processes:Whether a is a cause of b depends on facts aboutspatiotemporally restricted causal process, noton what would happen in other similar situa-tions. This means that theories that attempt todefine causality in terms of regularities (such asHume’s constant conjunction theory and manyprobabilistic theories of causation) are notcompatible with mechanism-based theories.

(Obviously, this does not mean that the sup-porters of the mechanism-based perspectivewould have to disregard regularities as animportant source of evidence about causalrelations.)

Some authors (e.g., Glennan 1996) haveattempted to define causation in terms ofmechanisms, but this approach faces sometricky problems. First, most characterizationsof mechanisms employ causal notions, so adefinition of causation in terms of mechanismswould be circular. Second, such an accountwould face a thorny question about the causalpowers of fundamental (physical) entities. Ifcausal relations at the fundamental level arenot mechanical, the definition is false, andif they are mechanical, we end up with aninfinite regress, which is regarded by many asan unhappy consequence.

Many supporters of mechanism-based ex-planations find Salmon’s (1984, 1998) con-served quantities theory of causal processesintuitively appealing. This is understandable,given that Salmon was one of the first philoso-phers who talked about the importance ofa causal-mechanical account of explanation.However, Salmon’s theory has turned out to bea disappointment. In addition to many technicaland philosophical problems associated with thetheory (Woodward 2003, pp. 350–58), it seemsquite ill suited to provide a foundational ac-count of a mechanism-based explanation. First,it is very difficult to see how the theory couldbe applied at all in biological and social sciences(or even some parts of physics), where explana-tions do not attempt to track spatiotemporallycontinuous processes such as transfer of energyor momentum. Second, as Hitchcock (1995) hasshown, Salmon’s approach is marred by the factthat it does not include any considerations ofexplanatory relevance. As a consequence, thesame counterexamples that have been raisedagainst the covering-law account can be raisedagainst Salmon’s approach (see the next sectionfor some examples).

A more natural complement to themechanism-based approach would seem to bethe account of causation recently developed by


Ann

u. R

ev. S

ocio

l. 20

10.3

6:49

-67.

Dow

nloa

ded

from

ww

w.a

nnua

lrev

iew

s.or

gby

Uni

vers

ity o

f C

alif

orni

a -

Sant

a C

ruz

on 0

9/16

/12.

For

per

sona

l use

onl

y.


Woodward (2002; 2003; see also Morgan &Winship 2007). In Woodward’s account (seeTable 1), causal claims track relations of coun-terfactual dependency. They tell us what wouldhave happened to the effect if the cause had beensubject to a surgical intervention that wouldnot have affected any other part of the causalstructure. One of the novelties of Woodward’stheory is its account of causal generalizations interms of invariances. According to Woodward,the explanatory qualities of a generalization aredetermined by its ability to tell us about thecounterfactual consequences of possible inter-ventions, not by any of the properties tradition-ally associated with laws of nature. Woodward’saccount supplements the mechanism-basedaccount by providing an account of explana-tory relevance and by making sense of causalproduction. However, it does not preemptthe importance of the mechanism-basedperspective. The mechanism approach differsfrom Woodward’s approach by its emphasis onthe importance of opening up black boxes andmaking explicit the causal cogs and wheelsthrough which effects are brought about.

The relevance of mechanisms is not limitedto explanation. Especially in nonexperimentalcontexts, mechanisms often have a crucial roleto play in distinguishing true causal relationsfrom spurious correlations. Mechanisms helpin causal inference in two ways. The knowl-edge that there is a mechanism through whichX influences Y supports the inference that X isa cause of Y. In addition, the absence of a plau-sible mechanism linking X to Y gives us a goodreason to be suspicious of the relation being acausal one (Hedstrom 2005). The knowledgeof mechanisms also has an important role in ex-trapolation of causal findings from one settingto another. As Steel (2008) discusses, the as-sumption about the similarity of causal mecha-nisms is crucial for making inferences from onesetting or population to another.

It is important to emphasize that mecha-nisms are not some sort of magic bullet forcausal inference, however. The problem oftenis not the absence of possible mechanisms, buthow to discriminate between a number of

potential mechanisms. To avoid lazymechanism-based storytelling, the mechanismscheme must be made explicit and detailed, andits assumptions must be supported by relevantempirical evidence.

Although it may be too strong to say that thespecification of mechanisms is always necessaryfor causal inference (Steel 2008), a fully satisfac-tory social scientific explanation requires thatthe causal mechanisms be specified. The pri-mary epistemic, in contrast to practical aim, ofscience is to understand phenomena, and thisis precisely what mechanisms provide. If wetake understanding to be an ability to answerwhat-if questions (Woodward 2003, Ylikoski2009, Ylikoski & Kuorikoski 2010), the con-tribution of the mechanism-based informationbecomes apparent. A simple causal claim tellsus about counterfactual dependency: It tells uswhat would have happened if the cause hadbeen different. The mechanism tells us whythe counterfactual dependency holds and tiesthe relata of the counterfactual to the knowl-edge about entities and relations underlying it.The account of a causal mechanism integratesan isolated piece of causal knowledge with amuch larger body of knowledge and helps us toanswer many natural follow-up questions aboutthe conditions under which the causal depen-dency holds: For example, what are the neces-sary background conditions and what are thepossible intervening factors that have to be ab-sent for the effect to be present? In this way themechanism expands our ability to answer what-if questions, i.e., it deepens our understanding(Ylikoski & Kuorikoski 2010).

A Comparison with theCovering-Law Account

The mechanism-based account of explanationin part has been developed as an alternativeto the once dominant covering-law accountof explanation of Hempel (1965) and others.The basic idea of the covering-law account isvery simple: An explanation is a deductive (orstatistical) argument that has a description ofthe explanandum phenomenon as a conclusion


Ann

u. R

ev. S

ocio

l. 20

10.3

6:49

-67.

Dow

nloa

ded

from

ww

w.a

nnua

lrev

iew

s.or

gby

Uni

vers

ity o

f C

alif

orni

a -

Sant

a C

ruz

on 0

9/16

/12.

For

per

sona

l use

onl

y.


and one or more empirically validated generallaw statements and a set of statements describ-ing particular facts (the initial conditions) asits premise. The core underlying idea is thatexplanations make the explanandum expected.This means that explanation and prediction aremore or less the same thing; the only differenceis that in the case of explanation we alreadyknow the outcome.

On the basis of the extensive debate that hastaken place in the philosophy of science duringthe past few decades (Salmon 1989, Woodward2003) and that has not yet been fully appreciatedby some social scientists (see, e.g., Opp 2005),we may safely conclude that the covering-lawapproach is a failure as a theory of explanation.Some of its failures are instructive, however,and it is useful to see how the mechanism-basedalternative avoids them.

The first set of problems concerns explana-tory relevance. The covering-law account hasbeen unable to make sense of the asymmetry ofexplanatory relations. Although effects do notexplain their causes, nothing in the covering-law account rules this out. As a consequence,according to the covering-law account, not onlycan the length of a flagpole explain the lengthof its shadow, the length of the shadow also canexplain the length of the flagpole (Salmon 1989,p. 103). In addition, the philosophical literatureis full of examples that satisfy all the require-ments of the covering-law account but that nev-ertheless are clearly nonexplanatory, such asthat the fact that a man consumes contracep-tive pills is not what explains why he does notget pregnant (Salmon 1989, p. 50; Woodward2003).

Another set of problems concerns the no-tion of law. The first covering-law account wasdeveloped with the expectation that the notionof a law could be given a satisfactory philo-sophical account, but this has not yet happened(Woodward 2003). In addition, there is an em-barrassing scarcity of covering laws; in sciencessuch as biology, psychology, and the social sci-ences, there are hardly any observable empiricalregularities that could be considered explana-tory (Cummins 2000).

The mechanism-based account is not af-fected by these problems. First, causality takescare of the problems of explanatory relevance:The direction of the explanation derives fromthe direction of causation (Salmon 1998), andthe counterfactual idea of difference-makinghelps to solve the problems related to irrel-evant explanatory factors (Woodward 2003).Similarly, the problems related to laws canbe ignored. The absence of covering lawsis not a problem, as they are not doing anyof the explanatory work. On the contrary,in mechanism-based accounts, observableregularities are the things to be explained(Cummins 2000). Of course, mechanism-basedexplanations still rely on causal generaliza-tions about the properties, activities, andrelations of underlying entities, but they donot have to satisfy the traditional criteria forlaws. Rather, it is sufficient that they fulfillWoodward’s (2003) invariance requirements.The mechanism-based account also seversthe close connection between explanationand prediction: It emphasizes the differencebetween diagnostic and explanatory reasoning,and although explanatory understanding isconstituted by an ability to make correct what-if inferences, this does not imply that it alwaysprovides a basis for empirical predictions.Finally, the mechanism-based theory of ex-planation is not wedded to the idea that anexplanation is a deductive argument (Ylikoski2005). The deductive reconstruction of theexplanation is an advisable practice for check-ing it for consistency, sufficiency, and implicitpremises, but explanation is not a relationbetween linguistic entities (Ylikoski 2007).

SOCIAL MECHANISMS

As mentioned in the introduction, over theyears the idea of mechanism-based explanationhas become increasingly influential in the so-cial sciences. It has proven to be a useful toolfor criticizing existing research practices andviews about the nature of the social scientificenterprise. Many different definitions of socialmechanism have been articulated, and quite


Ann

u. R

ev. S

ocio

l. 20

10.3

6:49

-67.

Dow

nloa

ded

from

ww

w.a

nnua

lrev

iew

s.or

gby

Uni

vers

ity o

f C

alif

orni

a -

Sant

a C

ruz

on 0

9/16

/12.

For

per

sona

l use

onl

y.


different methodological and theoretical con-clusions have been drawn (see Table 1 for someof the most influential definitions).

Raymond Boudon was an important earlycontributor to the mechanism approach. Insome of his publications from the 1970s, he ar-gued for the importance of “generative models”for explaining social outcomes (Boudon 1979;see also Fararo 1969), and he used computersimulations for this purpose (Boudon 1974). Ina famous exchange with Robert Hauser, one ofthe most vocal proponents of a strict statisticalapproach to sociological research during thatera, Boudon expressed one of the core ideas ofthe mechanism approach. He noted that statis-tical models of the sort advocated by Hauser areuseful for many purposes but that their useful-ness as causal explanations is considerably morerestricted than Hauser believed. Boudon em-phasized that explanations are not achieved bysimply estimating parameters of generic statis-tical models, but by developing generative mod-els that explicate the mechanisms at work (seeBoudon 1976, Hauser 1976).

Jon Elster has probably been the mostinfluential advocate of mechanisms in thesocial sciences, and his many books are full ofexcellent examples of mechanism-based think-ing in action. His idea that mechanism-basedexplanations open up black boxes and showthe cogs and wheels of the internal machinerycaptures quite well the mechanism-basedspirit. However, the various definitions ofmechanisms he has provided have been asource of some confusion (see Table 1).His early view (Elster 1989), according towhich a mechanism explains by providing acontinuous and contiguous chain of causal orintentional links between the explanans and theexplanandum, is quite in line with the generalmechanism-based perspective, although hischoice of wording wrongly suggests that inten-tional explanations are not causal explanations.In contrast, his later definition, “mechanismsare frequently occurring and easily recogniz-able causal patterns that are triggered undergenerally unknown conditions or with inde-terminate consequences” (Elster 1999, p. 1),

cannot serve as a general definition of mecha-nism. We should not require that the instancesof mechanisms occur frequently or that theybe easily recognizable, as sometimes findingunderlying mechanisms is the hardest part ofthe scientific work. Similarly, although therecertainly are cases in which we do not knowthe conditions that trigger the mechanism orother factors affecting the process, it would bea mistake to incorporate these conditions in thedefinition of a mechanism.

In historical sociology and political science,the late Charles Tilly was a prominent ad-vocate of the mechanism perspective. Despitehis inspiring empirical work, his general dis-cussions of mechanism-based explanation (e.g.,Tilly 2001) left something to be desired. Onegets a strong impression that he used the no-tion of a mechanism as a label to refer to thekind of processes that he for other reasons wasinterested in. Nevertheless, he had a clear ideaof how the mechanism perspective changes theresearch agenda in the historical study of poli-tics. Contentious episodes such as revolutions,nationalist mobilizations, and democratizationsdo not have essences. They are mostly the re-sult of similar kinds of social processes andthereby illustrate the fact that the same mecha-nisms can produce different outcomes in differ-ent circumstances. For this reason, comparativestudies should focus on these basic processesand mechanisms rather than on large-scaleepisodes whose classification is mostly a prod-uct of retrospective categorization (Tilly 2001,p. 36–38).

The critical realism movement (e.g., Sayer1992, Archer et al. 1998) has convinced many ofthe importance of mechanisms in social scien-tific explanations. The original source of theirideas about causation and scientific explana-tion is Harre’s (1970) pioneering work, butthe principal inspiration for critical realists hasbeen Bhaskar (1978, 1979). His critical realismis a tight philosophical package of ontologicaland epistemological views in which the idea ofmechanism-based explanation is only a small—but the most intuitive—part. Many social scien-tists rightly have been attracted by his criticism


Ann

u. R

ev. S

ocio

l. 20

10.3

6:49

-67.

Dow

nloa

ded

from

ww

w.a

nnua

lrev

iew

s.or

gby

Uni

vers

ity o

f C

alif

orni

a -

Sant

a C

ruz

on 0

9/16

/12.

For

per

sona

l use

onl

y.


of empiricist views of causation, but other partsof the package are problematic. For example,his transcendental argumentation, his layeredaccount of reality, and his ideas about essencesand internal relations have raised philosophicalsuspicions and doubts about their relevance forthe social sciences. The development of criti-cal realism also seems to have stalled: Althoughit is advocated in numerous books and articles,its supporters have largely tended to repeat orreformulate often quite cryptic original state-ments by Bhaskar instead of engaging with re-cent developments in the philosophy of scienceor using the ideas to explain important socialfacts.

Within philosophy of the social sciences,Daniel Little and Mario Bunge have beenimportant spokesmen for mechanism-basedthinking and, more generally, for scientificrealism. Little’s textbook (Little 1991) andarticles (Little 1998) show in an accessible wayhow thinking in terms of causal mechanismshelps to resolve various methodological prob-lems that characterize other approaches to thesocial sciences. His definition of mechanismas “a series of events governed by law-likeregularities that lead from the explanans to theexplanandum” (Little 1991, p. 15) is somewhatproblematic because of its reference to law-likeregularities, but it recognizes the importanceof causal generalizations in mechanisms, andLittle strongly emphasizes the important roleof microfoundations in mechanism-basedexplanations.

Bunge’s (1997, 2004) account ofmechanism-based explanation is part of hismore general systemistic philosophy of science.For him a system is “a complex object whoseparts or components are held together bybonds of some kind” (Bunge 2004, p. 188),and a mechanism is a set of processes that aretypical for material systems of certain kinds.These processes characterize the relationsand interactions between the system’s parts,its structure, and its environment. Just like inthe case of Bhaskar, Bunge’s basic ideas aboutmechanism-based explanations are in line withthe account outlined above, but they carry

extra philosophical baggage that we may wantto avoid.

In political science, causal mechanisms havehad an important role in debates concern-ing research methodologies and causal infer-ences (e.g., Mahoney 2001, George & Bennett2005, Gerring 2008, Box-Steffensmeier et al.2008). Mechanism-based thinking often hasbeen presented there as an alternative to sta-tistical methodologies and as providing founda-tions for case studies and other small-N studies.As George & Bennett (2005) show, the idea ofprocess tracing is useful for development andtesting of mechanism-based explanations, andit is not the case that only statistical evidence isrelevant for the assessment of causal claims. Ingeneral, the debate in political science has beenquite similar to that in sociology (see Sørensen1998): The advocates of mechanisms have crit-icized the simplistic empiricist uses of statisti-cal methodology for ignoring the importance ofcausal process assumptions in causal inference.

In criminology, the mechanism approachis particularly associated with the work of P-O Wikstrom. In opposition to the dominantrisk-factor paradigm within quantitative crim-inology, Wikstrom’s so-called situational ac-tion theory is a general theory of criminal be-havior that specifies a situational mechanismthrough which individual and situational factorsjointly influence the probability of individuals’engaging in criminal activities (see Wikstrom2006, Wikstrom & Treiber 2009). The situa-tional action theory is particularly concernedwith moral rules and the mechanisms that ex-plain why some people follow and others breachmoral rules as defined in law.

In addition to the above-mentioned schol-ars, Hedstrom & Swedberg’s (1998a) editedvolume Social Mechanisms includes many socialscientists who have been important in makingthe mechanism-based perspective visible. Thecontributions by Boudon, Elster, Gambetta,Hedstrom, Hernes, Schelling, and Sørensenrepresent what has become known as the an-alytical approach to social mechanisms, and wediscuss this approach in more detail below. Itwould be a mistake to take this collection of


Ann

u. R

ev. S

ocio

l. 20

10.3

6:49

-67.

Dow

nloa

ded

from

ww

w.a

nnua

lrev

iew

s.or

gby

Uni

vers

ity o

f C

alif

orni

a -

Sant

a C

ruz

on 0

9/16

/12.

For

per

sona

l use

onl

y.


essays as a definite formulation of this analyt-ically oriented approach, however. There areimportant differences between the authors, andthe approach is still developing (see in particu-lar Hedstrom 2005 and the various chapters inHedstrom & Bearman 2009a). In the following,we comment on some of these developments aswe try to articulate a more systematic accountof the implications of mechanism-based ideasfor the social sciences.

FROM MECHANISMS TOANALYTICAL SOCIOLOGY

Although the idea of mechanism-based expla-nation helps social scientists to avoid somephilosophical pitfalls, the mere adoption ofmechanism talk will not suffice. Much dependson how mechanism ideas are put to use; oth-erwise we end up with mere mechanism-basedstorytelling that lacks both theoretical rigor andempirical relevance. A broader vision of sociol-ogy is needed, which the so-called analytical so-ciology movement has attempted to articulate(see Hedstrom & Bearman 2009a). The iden-tity of this movement is not based on a commonobject of study, a shared historical tradition insociological theory, or the use of a specific re-search method or type of empirical data, buton some general ideas about what good socialscience is all about. At the most fundamentallevel, analytical sociology emphasizes the im-portance of the intellectual virtues of precisionand clarity. As Elster (2007, p. 455) expressedit: “What one might call the ‘analytical turn’in social science. . .rests. . .on a near-obsessiveconcern with clarity and explicitness.” That isto say, social scientific theories should be pre-sented with such precision and clarity that it ispossible to clearly distinguish the causal mech-anisms upon which they are founded.

The importance of the intellectual virtuesof clarity and explicitness derives in part fromour susceptibility to the illusion of depth of un-derstanding: We tend to overestimate the de-tail, coherence, and depth of our understanding(Keil 2003, Ylikoski 2009). With its implicit andvague explanatory standards, its largely verbal

mode of theorizing, and its highly complex ob-ject of study, social theory is probably a pri-mary example of this failure of meta-cognition.One of the key ways to fight this illusion is bymaking explanations explicit: Clearly and pre-cisely articulated accounts of mechanisms canbe subjected to a piecemeal scrutiny, and theirimplications can be assessed more accurately.

Analytical sociology emphasizes the im-portance of closely integrating theoretical andempirical work, but it is not closely tied to anyspecific research methodology. The crucialquestion is what kind of access a certain pieceof evidence provides to the causal process ofinterest, not whether the method being usedis quantitative or qualitative, experimental ornonexperimental.

Although the principles of analytical sociol-ogy are still developing, some central character-istics of the approach can be discerned. In thefollowing, we discuss four issues that are closelyrelated to the idea of mechanism-based expla-nation: (a) the typical sociological explananda,(b) the role of a theory of action, (c) mechanism-based theories as middle-range theories, and(d ) the role of agent-based modeling in ana-lytical sociology.

The Sociological Explananda

For sociology, the most important explanandaare social phenomena. Although sociologyshares with psychology and social psychologythe interest in explaining properties and behav-iors of individuals, its key challenge is to accountfor collective phenomena that are not definableby reference to any single member of the col-lectivity. Among such properties are (Hedstrom2005, p. 67)

1. Typical actions, beliefs, or desires amongthe members of society or a collectivity.

2. Distributions and aggregate patterns suchas spatial distributions and inequalities.

3. Topologies of networks that describerelationships between members of acollectivity.


Ann

u. R

ev. S

ocio

l. 20

10.3

6:49

-67.

Dow

nloa

ded

from

ww

w.a

nnua

lrev

iew

s.or

gby

Uni

vers

ity o

f C

alif

orni

a -

Sant

a C

ruz

on 0

9/16

/12.

For

per

sona

l use

onl

y.


4. Informal rules or social norms that con-strain the actions of the members of acollectivity.

From the point of view of a mechanism-based explanation, the basic entities that ex-plain these kinds of phenomena are humanagents and their relations. The so-called Cole-man’s boat illustrates clearly this aspect of themechanism-based approach (see Figure 1).

A basic point of the mechanism perspectiveis that explanations that simply relate macroproperties to each other (arrow 4) are unsatis-factory. These explanations do not specify thecausal mechanisms by which macro propertiesare related to each other. Deeper explanatoryunderstanding requires opening up the blackbox and finding the causal mechanisms thathave generated the macro-level observation.Rather than analyzing relationships betweenphenomena exclusively on the macro level, oneshould identify the situational mechanisms bywhich social structures constrain individuals’action and cultural environments shape theirdesires and beliefs (arrow 1), describe theaction-formation mechanisms linking individ-uals’ desires, beliefs, etc., to their actions (arrow2), and specify the transformational mech-anisms by which individuals, through theiractions and interactions, generate various in-tended and unintended social outcomes (arrow3). Only by understanding the whole chain ofsituational, action-formation, and transforma-tional mechanisms have we made sense of theobserved macro-level relationship (Coleman1990, Hedstrom & Swedberg 1998b).

This emphasis on detailing mechanisms im-plies that explanations should refer to individ-uals, their relationships, and their actions (inaddition to whatever other social properties arerelevant for explaining the causes and conse-quences of their actions). It is important to rec-ognize that the basic building blocks of socialexplanations are not mutually independent ac-tions performed by atomistic individuals, how-ever. Rather, individuals’ actions typically areoriented toward others, and their relations toothers therefore are central when it comes to

Situationalmechanisms

Transformationalmechanisms

Action-formationmechanisms

2

4

1 3

Macro-levelassociation

Figure 1A typology of social mechanisms (Hedstrom & Swedberg 1998b).

explaining why they do what they do. In addi-tion, social relations are central for explainingwhy, acting as they do, individuals bring aboutthe social outcomes they do. Consider, as anexample, the case of vacancy chains as analyzedby White (1970). A retirement, motivated per-haps by a desire for a more leisurely life, createsan opportunity for others, i.e., a job vacancywaiting to be filled by a new occupant. The va-cancy generated by the retirement is filled byanother individual whose reason for taking thejob, perhaps, is to attain more status, a highersalary, or just a change in venue, but this createsanother vacancy in this person’s old job, andin this way vacancy chains create social inter-dependencies that are important for explainingmobility. Individuals’ desires—for retirement,promotion, status, or change in venue—motivate the system. Without such orienta-tions, people may not move. But explanatoryunderstanding is only achieved by recognizingthat actions take place in relational structuresthat in this case channel mobility opportunitiesand thereby explain why we observe what weobserve (Hedstrom & Bearman 2009b).

The above argument does not imply anycommitment to the doctrine of methodologicalindividualism. Most formulations of method-ological individualism are much stronger thanthe mechanism-based perspective requires (fora review of the literature on methodologicalindividualism, see Udehn 2001). Some form


Ann

u. R

ev. S

ocio

l. 20

10.3

6:49

-67.

Dow

nloa

ded

from

ww

w.a

nnua

lrev

iew

s.or

gby

Uni

vers

ity o

f C

alif

orni

a -

Sant

a C

ruz

on 0

9/16

/12.

For

per

sona

l use

onl

y.


of structural individualism (Lindenberg 1977,Wippler 1978, Raub 1982, Coleman 1990,Udehn 2001) is sufficient for the purposes ofmechanism-based explanations of social phe-nomena. Structural individualism is a doctrineaccording to which all social facts, their struc-ture and change, are in principle explicable interms of individuals, their properties, actions,and relations to one another. Structural in-dividualism differs from most formulations ofmethodological individualism by emphasizingthe explanatory importance of relations andrelational structures. It does not require thatall explanatory facts are facts about individualagents in the strict sense: Facts about topologiesof social networks; about distributions of be-liefs, resources, or opportunities; and about in-stitutional or informal rules and norms can havea significant role in mechanism-based explana-tions. For example, the transformational mech-anisms (arrow 3 in Figure 1) are not based onsimple aggregation of individual actions, as inmany formulations of methodological individ-ualism, but depend on structural arrangementsthat cannot be defined by reference to atom-istic attributes of individual agents. Structuralindividualism is far from a trivial doctrine, how-ever. It is incompatible with radical holism andstructuralism because it requires that the socialpattern to be explained should be understoodin terms of interactions of individual agents.

Theories of Action

As action plays a central role in social mech-anisms, it raises the question concerning therole of a theory of action. The idea of a so-cial mechanism in itself does not tell us how toconceptualize human action. Rather than re-lying on some preconceived ideas about hu-man motivation or cognitive processing, themechanism-based perspective suggests that ouraccount of human agency should be based onfindings and theories of psychological and cog-nitive sciences.

The explanatory agenda of mechanism-based social science has some methodologi-cal implications, however. As understanding of

complex phenomena is only possible in a piece-meal way, we must abstract away from many de-tails of human mental life. Only those aspects ofcognition that are relevant for the explanatorytask at hand should be included in the explana-tion, and the explanatory task thus determineshow rich the psychological assumptions mustbe.

So although the mechanism-based approachemphasizes the importance of action in theexplanation of social phenomena, it does notsubscribe to an axiomatic vision accordingto which a specific action theory should beused for all purposes. For many social scien-tific purposes, a relatively simple desire-belief-opportunity model will be sufficient (Hedstrom2005). For other purposes, a pragmatist the-ory of action (Gross 2009) could be fruitful.Habits, routines, and various unconscious cog-nitive processes are important parts of mod-ern naturalistic accounts of human cognition(Wilson 2002, Hassin et al. 2005), and sociologyneeds to take such factors into account. Theirincorporation into detailed sociological modelsrequires much additional work, however.

The idea of social mechanisms is quiteoften associated with rational choice theory(Abbott 2007, Gross 2009). However, froma philosophical and general sociological pointof view, the connection between the two isquite weak. There is nothing in the idea of amechanism-based explanation that would re-quire the explanation to be articulated in termsof rational choice theory. On the contrary,the requirement that mechanism-based expla-nations cite actual causes of the phenomenonto be explained often makes rational choiceexplanations unacceptable, as they are builtupon implausible psychological and sociolog-ical assumptions. Empirically false assumptionsabout human motivation, cognitive processes,access to information, or social relations cannotbear the explanatory burden in a mechanism-based explanation. It is not enough that themodel “saves the phenomena”; it should rep-resent the essential features of the actual causalstructure that produces the observed phenom-ena. Rather than accepting the instrumentalist


Ann

u. R

ev. S

ocio

l. 20

10.3

6:49

-67.

Dow

nloa

ded

from

ww

w.a

nnua

lrev

iew

s.or

gby

Uni

vers

ity o

f C

alif

orni

a -

Sant

a C

ruz

on 0

9/16

/12.

For

per

sona

l use

onl

y.


“as if ” attitude displayed by many economists,the mechanism-based approach requires thatone should strive for theoretical assumptionsthat are empirically valid and compatible withthe results of other disciplines. In order for un-realistic psychological and social assumptions tobe acceptable, they must be simplifying ideal-izations that help the modeling but do not affectthe central explanatory relationships in any cru-cial manner. This rarely is the case in rationalchoice theory.

Mechanisms andMiddle-Range Theories

The idea of causal mechanisms is related tobroader ideas about the growth and organiza-tion of scientific knowledge. According to anold but still influential empiricist view, gen-eral scientific knowledge consists of empiricalgeneralizations and more abstract theoreticalprinciples from which these generalizations can(ideally) be deduced. This vision is challengedby the mechanism-based account of knowledge,with its emphasis on scientific knowledge as em-bedded in mechanism schemes and not in em-pirical generalizations.

According to this view, scientific knowledgeexpands by adding items to or improving uponitems already present in the toolbox of possi-ble causal mechanisms. Understanding accu-mulates as the knowledge of mechanisms getsmore detailed and the number of known mech-anisms increases. This vision of knowledge doesnot require that mechanisms be ultimately or-ganized into a grand unified theory. It is onlyrequired that the accounts of mechanisms pro-vided by different disciplines be mutually com-patible and that they form an integrated web inwhich mechanisms at lower levels of organiza-tion explain the mechanisms that higher-leveldisciplines take for granted. For example, psy-chology explains (and corrects) the assumptionsthat social scientists make about human cogni-tive processes.

The mechanism idea is important in a highlyspecialized and fragmented discipline such associology. Although empirical data, research

methods, and substantial theories differ fromone subfield of sociology to another, the gen-eral ideas about possible causal mechanisms aresomething these fields could share and therebybenefit from each other’s work. In this vision,sociological theory provides a set of explana-tory tools that can be employed and adaptedto particular situations and explanatory tasks.The mechanisms are (semi) general in the sensethat most of them are not limited to any par-ticular application. For example, the same typeof mechanism can be used for (partially) ex-plaining residential segregation (Bruch & Mare2006) and success in cultural markets (Salganik& Watts 2009).

This mechanism-based vision of knowl-edge has much in common with Merton’s ideaof sociological theories of the middle range(Merton 1968; see also Hedstrom & Udehn2009). According to Merton (1968, p. 39), the-ories of the middle range are

theories that lie between the minor butnecessary working hypotheses that evolvein abundance during day-to-day researchand the all-inclusive systematic efforts todevelop a unified theory that will explain allthe observed uniformities of social behavior,social organization and social change.

An ideal theory of the middle range is clear,precise, and simple. It does not bore the readerby attempting to describe the causal process inall its detail; instead it seeks to highlight theheart of the story by isolating a few explanatoryfactors that explain important but delimited as-pects of the outcomes to be explained. A theoryof the middle range can be used for partially ex-plaining a range of different phenomena, but itmakes no pretense of being able to explain allsocial phenomena, and it is not founded uponany form of extreme reductionism.

Merton’s account of self-fulfilling prophe-cies provides a clear example of a social mecha-nism. The focus of the analysis is on the processthrough which an initially false belief evokesbehavior that eventually makes the false be-lief come true, and its basic idea is captured by


Ann

u. R

ev. S

ocio

l. 20

10.3

6:49

-67.

Dow

nloa

ded

from

ww

w.a

nnua

lrev

iew

s.or

gby

Uni

vers

ity o

f C

alif

orni

a -

Sant

a C

ruz

on 0

9/16

/12.

For

per

sona

l use

onl

y.


the so-called Thomas Theorem: “If men definesituations as real, they are real in their conse-quences” (Merton 1968, p. 475). Merton showshow an endogenous and self-reinforcing pro-cess can bring about a collective outcome that isunintended by all the individuals involved, andthe canonical example he uses is a run on a bank.Once a rumor of insolvency gets started, somedepositors are likely to withdraw their savings,acting on the principle that it is better to besafe than sorry. Their withdrawals strengthenthe beliefs of others that the bank is in financialdifficulties, partly because the withdrawals mayactually hurt the financial standing of the bank,but more importantly because the act of with-drawal in itself signals to others that somethingmight be wrong with the bank. This produceseven more withdrawals, which further strength-ens the belief, and so on. By this mechanism,even an initially sound bank may go bankrupt ifenough depositors withdraw their money in the(initially) false belief that the bank is insolvent.

The mechanism sketched by Merton is animportant part of the theoretical toolbox of so-ciology, and it is an ideal example of what theanalytically oriented mechanism approach is allabout. The basic structure of Merton’s mech-anism is as depicted in Figure 2 (Hedstrom &Udehn 2009) and it has the following generalargument pattern: The beliefs of one or sev-eral Alters are such that they decide to performa certain action; their actions influence Ego’sbelief about the value of performing the act insuch a way that Ego also decides to act; andEgo’s action in turn strengthens the beliefs ofothers in the value of performing the act. Thistype of self-reinforcing and belief-centered cy-cle is then repeated and can make an initially

BA AA BE AE

Figure 2The logic of self-fulfilling prophecies. Thesubscripts identify the individuals in question, Egoand Alter(s), and B and A represent the beliefs andactions of the relevant individuals.

false belief come true. This mechanism is, inprinciple, quite general, and it can be used toexplain a range of different types of phenomena(Biggs 2009). The development of this mech-anism scheme also provides a good example ofgrowth of mechanism-based knowledge. Theoriginal sketch of a mechanism for a limitednumber of phenomena has grown to a wholefamily of models with a wide range of applica-tions.

Agent-Based Modeling

In their effort to develop clear and precise ex-planations of social phenomena, agent-basedcomputer simulation has emerged as an es-sential tool for analytical sociology (see Macy& Flache 2009). Much of the development ofmechanism-based knowledge consists of de-veloping how-possibly explanation schemes.These schemes are not intended to explain anyparticular empirical facts directly, but to pro-vide a general understanding of how thingscould work. Given the limitations of experi-mental methods and the complexity of socialphenomena, computer simulations are impor-tant for this kind of endeavor. Computer sim-ulations allow systematic exploration of con-sequences of modeling assumptions and makeit possible to model much more complex phe-nomena than was possible earlier.

The promise of agent-based simulation isbased on the fact that the dynamics observedat the social level typically are complex andhard to understand, but often it is possibleto describe the basic cogs and wheels ofthese social processes with rather simplemodels. Schelling’s (1978) famous segregationmodel is a good illustration of this. Schellingdemonstrated that even if individuals donot have strong preferences against mixedneighborhoods, segregation often arises asan unintended consequence. He also showedthat small and seemingly trivial differences atthe micro level can make huge differences atthe macro level. More generally, Schelling’ssimple and stylized example demonstrates thatmacro-level outcomes and relationships tell us


Ann

u. R

ev. S

ocio

l. 20

10.3

6:49

-67.

Dow

nloa

ded

from

ww

w.a

nnua

lrev

iew

s.or

gby

Uni

vers

ity o

f C

alif

orni

a -

Sant

a C

ruz

on 0

9/16

/12.

For

per

sona

l use

onl

y.


very little about why we observe the macro-level outcomes and relationships we observe.Only by explicitly considering the micro-levelactions and relations and how they unfold overtime can macro-level outcomes be explained.This basic insight is at the heart of analyticalsociology: To understand collective dynamics,we must study the collectivity as a whole, butwe must not study it as a collective entity.

One important feature of agent-basedsimulations is that they do not impose any apriori constraints on the mechanisms assumedto be operating. Unlike rational choice theory,agent-based modeling is not based on anyspecific theory of action or interaction. It is amethodology for deriving the social outcomesthat groups of interacting actors are likelyto bring about, whatever the action logics orinteraction structures may be. The simulationallows the researcher to see how the phe-nomenon to be explained is generated and howchanges in action logics or relational structuresare likely to change the social outcome.

As emphasized by Manzo (2007), agent-based modeling is not only a useful tool foranalytical sociology, but there is a natural affin-ity between the components of mechanism-based explanations and agent-based simulationmodels:

When we write a set of computational algo-rithms (the program), formalizing the gen-erative hypotheses. . . , what we are doing ishypothesizing a series of generative mecha-nisms. When we execute the program . . . weengender the process deriving from the setof posited generative mechanisms. With thetechnical distinction between program “writ-ing,” “compilation,” and “execution” it be-comes clear that a “process” is nothing morethan the dynamic aspect of one (or several)mechanism(s): it is what the mechanism cantrigger (Manzo 2007, pp. 5–6).

That is to say, agent-based simulation methodsprovide a technical infrastructure tightlycoupled to the theoretical agenda of analyticalsociology. The simulation allows us to see how

the phenomenon to be explained is generatedand how changes in the assumptions change theoutcome. If explanatory understanding consistsof knowledge of dependencies and ability tomake correct what-if inferences on the basis ofthis knowledge (Ylikoski & Kuorikoski 2010),it is clear that computer simulations increaseour ability to see the dependencies that followfrom different sets of assumptions and that itincreases the reliability of our inferences. Aswas so eloquently illustrated by Schelling inhis segregation analysis, the latter is especiallyimportant, as our unaided inferences are quiteerror-prone.

Agent-based simulations should not only beregarded as a tool for theoretical exploration,however. Empirically calibrated agent-basedmodels make it possible to integrate theoreticalideas with the results of empirical research.Simulation studies that aim to account forempirical observations employ a generativeresearch strategy (Hedstrom & Bearman2009b, Epstein 2006). The basic structure ofthis research strategy is the following:

1. Start with a clearly delineated social factthat is to be explained.

2. Formulate different hypotheses about rel-evant micro-level mechanisms.

3. Translate the theoretical hypotheses intocomputational models.

4. Simulate the models to derive the type ofsocial facts that each micro-level mecha-nism brings about.

5. Compare the social facts generated byeach model with the actually observedoutcomes.

One recent study that nicely illustratesthis approach is the study undertaken byBearman et al. (2004) of the sexual and roman-tic networks of adolescents. The context oftheir study was a single high school of roughly1000 students in the Midwestern UnitedStates. The macro structure they sought toexplain was the surprising discovery that thesexual and romantic network structure of thestudents resembled a spanning tree. Theyidentified different micro-level processes thatcould explain this macro-level pattern and


Ann

u. R

ev. S

ocio

l. 20

10.3

6:49

-67.

Dow

nloa

ded

from

ww

w.a

nnua

lrev

iew

s.or

gby

Uni

vers

ity o

f C

alif

orni

a -

Sant

a C

ruz

on 0

9/16

/12.

For

per

sona

l use

onl

y.


used simulations to derive what the macrostructure would look like if a particular microprocess were at work. By performing differentsimulations, they came to the conclusion thatthe spanning-tree structure most likely wasthe result of a social norm that prohibitedcycles of length four. From a boy’s point ofview, this norm implies that he should notform a partnership with his prior girlfriend’scurrent boyfriend’s prior girlfriend. Bearmanet al. convincingly argued that the status lossassociated with such relationships is a majorfactor enforcing this norm, as is the so-calledyuck factor that arises from too much closeness.

CONCLUSIONS

During the last decade, mechanisms andmechanism-based explanations have attracteda great deal of attention in the social sciencesas well as in the philosophy of science. Inthis article, we have reviewed the most im-portant mechanism-related literature and dis-cussed how ideas about mechanism-based ex-planations fit into a broader view of the aimsand standards of sociology.

Much has been accomplished during theseyears, and the literature on mechanisms con-tinues to be intellectually exciting and to fo-cus on issues of considerable relevance for thefuture development of the discipline. In theseconcluding remarks, we reconnect to a themethat we have touched upon above: the dangersof lazy mechanism-based storytelling.

If the mechanism approach is to deliver onits promise, it must be taken seriously. Con-tinuing along the same paths as before andsimply interpreting one’s research findings inmechanism terms will not suffice. A propermechanism-based explanation is not an ad hocaddition that can be added to any set of em-pirical observations at will. Underlying themechanism-based approach is a commitmentto realism and an opposition to any form of in-strumentalism: The explanations should reflectthe causal processes actually responsible for theobservations.

The mechanism-based approach requiresstringency in theoretical practice and imagina-tion in the design of empirical research. Thereis still much work to be done in the transla-tion of the general principles of the mechanismapproach into concrete research practices, butwe find recent developments in analytical so-ciology promising in this regard. The key tofuture progress is the development of good ex-emplars of analytical sociology rather than, say,engaging in further debates about the propermechanism definition.

One of the basic premises of analytical so-ciology is that proper understanding of col-lective processes requires that the researcherspay attention to the entities that mechanismsare made of (the agents, their properties, ac-tions, and relations) rather than treating themas opaque entities. This dogma of analytical so-ciology means that theoretical models as wellas empirical research must focus on actions andrelations and how they unfold over time. Thisbasic principle has far-reaching consequencesfor the type of empirical data that needs to beused as well as for the type of theoretical workthat is needed.

As far as data and empirical research are con-cerned, longitudinal data with relational infor-mation are essential given that the basic en-tities of social mechanisms are actors, theiractions, and relations. Similarly, the mecha-nism approach has considerable consequencesfor how theoretical work ought to be con-ducted. As discussed above, in order to ex-plain macro-level change, rigorous theorizingis needed that explicitly considers the micro-level mechanisms at work and the dynamicprocesses that they give rise to. We are con-vinced that agent-based computer simulationswill become a basic tool of social theory ingeneral and mechanism-based theory in par-ticular. The flexibility of simulation models al-lows theoretical ideas to be translated into gen-erative models without distorting the originalideas. Although much work remains to be done,a further development of the mechanism ap-proach clearly calls for empirically calibrated


Ann

u. R

ev. S

ocio

l. 20

10.3

6:49

-67.

Dow

nloa

ded

from

ww

w.a

nnua

lrev

iew

s.or

gby

Uni

vers

ity o

f C

alif

orni

a -

Sant

a C

ruz

on 0

9/16

/12.

For

per

sona

l use

onl

y.


simulation models of the sort discussed above.Empirically calibrated simulation models seemto be the natural way of bridging the unfor-

tunate gap between theoretical and empiricalwork that has hampered the discipline for somany years.

DISCLOSURE STATEMENT

The authors are not aware of any affiliations, memberships, funding, or financial holdings thatmight be perceived as affecting the objectivity of this review.

ACKNOWLEDGMENTS

We thank Peter Bearman, Karen Cook, John Goldthorpe, Gianluca Manzo, and P-O Wikstromfor their valuable comments.

LITERATURE CITED

Abbott A. 2007. Mechanisms and relations. Sociologica 2:1–22Archer MS, Bhaskar R, Collier A, Lawson T, Norrie A, eds. 1998. Critical Realism: Essential Readings. London:

RoutledgeBearman P, Moody J, Stovel K. 2004. Chains of affection: the structure of adolescent romantic and sexual

networks. Am. J. Sociol. 110:44–91Bechtel W. 2006. Discovering Cell Mechanisms. The Creation of Modern Cell Biology. New York: Cambridge

Univ. PressBechtel W. 2008. Mental Mechanism. Philosophical Perspectives on Cognitive Neuroscience. London: RoutledgeBechtel W, Abrahamsen A. 2005. Explanation: a mechanist alternative. Stud. Hist. Philos. Biol. Biomed. Sci.

36:421–41Bechtel W, Richardson RC. 1993. Discovering Complexity. Decomposition and Localization as Strategies in Scientific

Research. Princeton, NJ: Princeton Univ. PressBhaskar R. 1978. A Realist Theory of Science. Brighton: Harvester. 2nd ed.Bhaskar R. 1979. The Possibility of Naturalism. Brighton: HarvesterBiggs M. 2009. Self-fulfilling prophecies. See Hedstrom & Bearman 2009a, pp. 294–314Boudon R. 1974. Education, Opportunity, and Social Inequality: Changing Prospects in Western Society. New York:

WileyBoudon R. 1976. Comment on Hauser’s “Review of Education, Opportunity, and Social Inequality.” Am. J.

Sociol. 81:1175–87Boudon R. 1979. Generating models as a research strategy. In Qualitative and Quantitative Social Research:

Papers in Honor of Paul F. Lazarsfeld, ed. PH Rossi, pp. 51–64. New York: Free PressBox-Steffensmeier JM, Brady HE, Collier D, eds. 2008. The Oxford Handbook of Political Methodology. Oxford:

Oxford Univ. PressBruch EE, Mare RD. 2006. Neighborhood choice and neighborhood change. Am. J. Sociol. 112:667–709Bunge M. 1997. Mechanism and explanation. Philos. Soc. Sci. 27:410–65Bunge M. 2004. How does it work? The search for explanatory mechanisms. Philos. Soc. Sci. 34:182–210Coleman JS. 1990. Foundations of Social Theory. Cambridge, MA: BelknapCraver C. 2007. Explaining the Brain. Mechanisms and the Mosaic Unity of Neuroscience. Oxford: Oxford Univ.

PressCummins R. 2000. “How does it work?” versus “What are the laws?”: two conceptions of psychological

explanation. In Explanation and Cognition, ed. FC Keil, RA Wilson, pp. 117–44. Cambridge, MA: MITPress

Darden L. 2006. Reasoning in Biological Discoveries. Essays on Mechanisms, Interfield Relations, and AnomalyResolution. Cambridge, UK: Cambridge Univ. Press

Elster J. 1989. Nuts and Bolts for the Social Sciences. Cambridge, UK: Cambridge Univ. Press


Ann

u. R

ev. S

ocio

l. 20

10.3

6:49

-67.

Dow

nloa

ded

from

ww

w.a

nnua

lrev

iew

s.or

gby

Uni

vers

ity o

f C

alif

orni

a -

Sant

a C

ruz

on 0

9/16

/12.

For

per

sona

l use

onl

y.


Elster J. 1999. Alchemies of the Mind: Rationality and the Emotions. Cambridge, UK: Cambridge Univ. PressElster J. 2007. Explaining Social Behavior: More Nuts and Bolts for the Social Sciences. Cambridge, UK: Cambridge

Univ. PressEpstein J. 2006. Generative Social Science. Studies in Agent-Based Computational Modeling. Princeton, NJ:

Princeton Univ. PressFararo TJ. 1969. The nature of mathematical sociology. Soc. Res. 36(1):75–92George AL, Bennett A. 2005. Case Studies and Theory Development in the Social Sciences. Cambridge, MA: MIT

PressGerring J. 2008. The mechanismic worldview: thinking inside the box. Br. J. Polit. Sci. 38:161–79Glennan S. 1996. Mechanisms and the nature of causation. Erkenntnis 44:49–71Glennan S. 2002. Rethinking mechanistic explanation. Philos. Sci. 69:S342–53Gross N. 2009. A pragmatist theory of social mechanisms. Am. Sociol. Rev. 74:358–79Harre R. 1970. The Principles of Scientific Thinking. London: MacmillanHassin RR, Uleman JS, Bargh JA. 2005. The New Unconscious. Oxford: Oxford Univ. PressHauser RM. 1976. Review essay: on Boudon’s model of social mobility. Am. J. Sociol. 81:911–28Hedstrom P. 2005. Dissecting the Social. On the Principles of Analytical Sociology. Cambridge, UK: Cambridge

Univ. PressHedstrom P, Bearman P, eds. 2009a. The Oxford Handbook of Analytical Sociology. Oxford: Oxford Univ. PressHedstrom P, Bearman P. 2009b. What is analytical sociology all about? An introductory essay. See Hedstrom

& Bearman 2009a, pp. 3–24Hedstrom P, Swedberg R, eds. 1998a. Social Mechanisms. An Analytical Approach to Social Theory. Cambridge,

UK: Cambridge Univ. PressHedstrom P, Swedberg R. 1998b. Social mechanisms: an introductory essay. See Hedstrom & Swedberg

1998a, pp. 1–31Hedstrom P, Udehn L. 2009. Analytical sociology and theories of the middle range. See Hedstrom & Bearman

2009a, pp. 25–49Hempel C. 1965. Aspects of Scientific Explanation. New York: Free PressHitchcock C. 1995. Discussion: Salmon on explanatory relevance. Philos. Sci. 62:304–20Keil FC. 2003. Folkscience: coarse interpretations of a complex reality. Trends Cogn. Sci. 7:368–73Kincaid H. 1996. Philosophical Foundations of the Social Sciences. Analyzing Controversies in Social Research.

Cambridge, UK: Cambridge Univ. PressKuorikoski J. 2009. Two concepts of mechanism: componential causal system and abstract form of interaction.

Int. Stud. Philos. Sci. 23:143–60Lindenberg S. 1977. Individuelle Effekte, Kollektive Phanomene und das Problem der Transformation. In

Probleme der Erklarung Sozialen Verhaltens, ed. K Eichner, W Habermehl, pp. 46–84. Meisenheim: AntonHain

Little D. 1991. Varieties of Social Explanation: An Introduction to the Philosophy of Social Science. Boulder, CO:Westview

Little D. 1998. Microfoundations, Method, and Causation. On the Philosophy of the Social Science. New Brunswick,NJ: Transaction

Machamer P, Darden L, Craver C. 2000. Thinking about mechanisms. Philos. Sci. 67:1–25Macy M, Flache A. 2009. Social dynamics from the bottom up: agent-based models of social interaction. See

Hedstrom & Bearman 2009a, pp. 245–68Mahoney J. 2001. Beyond correlational analysis: recent innovations in theory and method. Sociol. Forum

16(3):575–93Manicas P. 2006. A Realist Philosophy of Social Science. Explanation and Understanding. Cambridge, UK:

Cambridge Univ. PressManzo G. 2007. Comment on Andrew Abbott. Sociologica 2:1–8Mayntz R. 2004. Mechanisms in the analysis of social macro-phenomena. Philos. Soc. Sci. 34:237–59Merton RK. 1968. Social Theory and Social Structure. New York: Free PressMilgram S. 1967. The small world problem. Psychol. Today 2:60–67Morgan SL, Winship C. 2007. Counterfactuals and Causal Inference. Methods and Principles for Social Research.

New York: Cambridge Univ. Press


Ann

u. R

ev. S

ocio

l. 20

10.3

6:49

-67.

Dow

nloa

ded

from

ww

w.a

nnua

lrev

iew

s.or

gby

Uni

vers

ity o

f C

alif

orni

a -

Sant

a C

ruz

on 0

9/16

/12.

For

per

sona

l use

onl

y.


Norkus Z. 2005. Mechanisms as miracle makers? The rise and inconsistencies of the “mechanismic approach”in social science and history. Hist. Theory 44(3):348–72

Opp K-D. 2005. Explanations by mechanisms in the social sciences. Problems, advantages and alternatives.Mind Soc. 4:163–78

Raub W. 1982. The structural-individualist approach: towards an explanatory sociology. In Theoretical Modelsand Empirical Analyses, ed. W Raub, pp. 3–40. Utrecht: E.S. Publications

Salganik MJ, Watts DJ. 2009. Social influence: the puzzling nature of success in cultural markets. See Hedstrom& Bearman 2009a, pp. 315–41

Salmon W. 1984. Scientific Explanation and the Causal Structure of the World. Princeton, NJ: Princeton Univ.Press

Salmon W. 1989. Four Decades of Scientific Explanation. Minneapolis: Univ. Minn. PressSalmon W. 1998. Causality and Explanation. Oxford: Oxford Univ. PressSayer A. 1992. Method in Social Science. A Realist Approach. London: Routledge. 2nd ed.Schelling TC. 1978. Micromotives and Macrobehavior. New York: W.W. NortonSchmidt M. 2006. Die Logik mechanismischer Erklarungen. Wiesbaden: VS VerlagSkipper RA, Millstein RL. 2005. Thinking about evolutionary mechanisms: natural selection. Stud. Hist. Philos.

Biol. Biomed. Sci. 36:327–47Sørensen AB. 1998. Theoretical mechanisms and the empirical study of social processes. See Hedstrom &

Swedberg 1998a, pp. 238–66Steel D. 2008. Across the Boundaries. Extrapolation in Biology and Social Sciences. Oxford: Oxford Univ. PressThagard P. 1999. How Scientists Explain Disease. Princeton, NJ: Princeton Univ. PressTilly C. 2001. Mechanisms in political processes. Annu. Rev. Polit. Sci. 4:21–41Udehn L. 2001. Methodological Individualism: Background, History and Meaning. London: RoutledgeWatts DJ, Strogatz SH. 1998. Collective dynamics of ‘small world’ networks. Nature 393:440–42White HC. 1970. Chains of Opportunity: System Models of Mobility in Organizations. Cambridge, MA: Harvard

Univ. PressWikstrom P-O. 2006. Individuals, settings and acts of crime. Situational mechanisms and the explanation of

crime. In The Explanation of Crime: Context, Mechanisms and Development, ed. P-O Wikstrom, RJ Sampson,pp. 61–107. Cambridge, UK: Cambridge Univ. Press

Wikstrom P-O, Treiber K. 2009. Violence as situational action. Int. J. Confl. Violence 3:75–96Wilson TD. 2002. Strangers to Ourselves. Discovering the Adaptive Unconscious. Cambridge, MA: BelknapWimsatt W. 2007. Re-Engineering Philosophy for Limited Beings. Piecewise Approximations to Reality. Cambridge,

MA: Harvard Univ. PressWippler R. 1978. The structural-individualist approach in Dutch sociology: toward an explanatory social

science. Neth. J. Sociol. 14:135–55Woodward J. 2002. What is a mechanism? A counterfactual account. Philos. Sci. 69:S366–S377Woodward J. 2003. Making Things Happen. A Theory of Causal Explanation. Oxford: Oxford Univ. PressYlikoski P. 2005. The third dogma revisited. Found. Sci. 10:395–419Ylikoski P. 2007. The idea of contrastive explanandum. In Rethinking Explanation, ed. J Persson, P Ylikoski,

pp. 27–42. Dordrecht: SpringerYlikoski P. 2009. Illusions in scientific understanding. In Scientific Understanding: Philosophical Perspectives, ed.

H De Regt, S Leonelli, K Eigner, pp. 100–19. Pittsburgh, PA: Pittsburgh Univ. PressYlikoski P. 2010. Social mechanisms and explanatory relevance. In From Social Mechanisms to Analytical Sociology,

ed. P Demeulenaere. Cambridge, UK: Cambridge Univ. Press. In pressYlikoski P, Kuorikoski J. 2010. Dissecting explanatory power. Philos. Stud. 148:201–19


Ann

u. R

ev. S

ocio

l. 20

10.3

6:49

-67.

Dow

nloa

ded

from

ww

w.a

nnua

lrev

iew

s.or

gby

Uni

vers

ity o

f C

alif

orni

a -

Sant

a C

ruz

on 0

9/16

/12.

For

per

sona

l use

onl

y.

SO36-FM ARI 2 June 2010 4:38

Annual Reviewof Sociology

Volume 36, 2010Contents

FrontispieceJohn W. Meyer � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � xiv

Prefatory Chapter

World Society, Institutional Theories, and the ActorJohn W. Meyer � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 1

Theory and Methods

Causal Inference in Sociological ResearchMarkus Gangl � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �21

Causal Mechanisms in the Social SciencesPeter Hedstrom and Petri Ylikoski � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �49

Social Processes

A World of Standards but not a Standard World: Toward a Sociologyof Standards and StandardizationStefan Timmermans and Steven Epstein � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �69

Dynamics of Dyads in Social Networks: Assortative, Relational,and Proximity MechanismsMark T. Rivera, Sara B. Soderstrom, and Brian Uzzi � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �91

From the Sociology of Intellectuals to the Sociology of InterventionsGil Eyal and Larissa Buchholz � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 117

Social Relationships and Health Behavior Across the Life CourseDebra Umberson, Robert Crosnoe, and Corinne Reczek � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 139

Partiality of Memberships in Categories and AudiencesMichael T. Hannan � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 159

v

Ann

u. R

ev. S

ocio

l. 20

10.3

6:49

-67.

Dow

nloa

ded

from

ww

w.a

nnua

lrev

iew

s.or

gby

Uni

vers

ity o

f C

alif

orni

a -

Sant

a C

ruz

on 0

9/16

/12.

For

per

sona

l use

onl

y.

SO36-FM ARI 2 June 2010 4:38

Institutions and Culture

What Is Sociological about Music?William G. Roy and Timothy J. Dowd � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 183

Cultural Holes: Beyond Relationality in Social Networks and CultureMark A. Pachucki and Ronald L. Breiger � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 205

Formal Organizations

Organizational Approaches to Inequality: Inertia, Relative Power,and EnvironmentsKevin Stainback, Donald Tomaskovic-Devey, and Sheryl Skaggs � � � � � � � � � � � � � � � � � � � � � � � � 225

Political and Economic Sociology

The Contentiousness of Markets: Politics, Social Movements,and Institutional Change in MarketsBrayden G King and Nicholas A. Pearce � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 249

Conservative and Right-Wing MovementsKathleen M. Blee and Kimberly A. Creasap � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 269

The Political Consequences of Social MovementsEdwin Amenta, Neal Caren, Elizabeth Chiarello, and Yang Su � � � � � � � � � � � � � � � � � � � � � � � � � � 287

Comparative Analyses of Public Attitudes Toward Immigrantsand Immigration Using Multinational Survey Data: A Reviewof Theories and ResearchAlin M. Ceobanu and Xavier Escandell � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 309

Differentiation and Stratification

Income Inequality: New Trends and Research DirectionsLeslie McCall and Christine Percheski � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 329

Socioeconomic Disparities in Health BehaviorsFred C. Pampel, Patrick M. Krueger, and Justin T. Denney � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 349

Gender and Health InequalityJen’nan Ghazal Read and Bridget K. Gorman � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 371

Incarceration and StratificationSara Wakefield and Christopher Uggen � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 387

Achievement Inequality and the Institutional Structure of EducationalSystems: A Comparative PerspectiveHerman G. Van de Werfhorst and Jonathan J.B. Mijs � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 407

vi Contents

Ann

u. R

ev. S

ocio

l. 20

10.3

6:49

-67.

Dow

nloa

ded

from

ww

w.a

nnua

lrev

iew

s.or

gby

Uni

vers

ity o

f C

alif

orni

a -

Sant

a C

ruz

on 0

9/16

/12.

For

per

sona

l use

onl

y.

SO36-FM ARI 2 June 2010 4:38

Historical Studies of Social Mobility and StratificationMarco H.D. van Leeuwen and Ineke Maas � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 429

Individual and Society

Race and TrustSandra Susan Smith � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 453

Three Faces of IdentityTimothy J. Owens, Dawn T. Robinson, and Lynn Smith-Lovin � � � � � � � � � � � � � � � � � � � � � � � � � � 477

Policy

The New Homelessness RevisitedBarrett A. Lee, Kimberly A. Tyler, and James D. Wright � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 501

The Decline of Cash Welfare and Implications for Social Policyand PovertySandra K. Danziger � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 523

Indexes

Cumulative Index of Contributing Authors, Volumes 27–36 � � � � � � � � � � � � � � � � � � � � � � � � � � � 547

Cumulative Index of Chapter Titles, Volumes 27–36 � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 551

Errata

An online log of corrections to Annual Review of Sociology articles may be found athttp://soc.annualreviews.org/errata.shtml

Contents vii

Ann

u. R

ev. S

ocio

l. 20

10.3

6:49

-67.

Dow

nloa

ded

from

ww

w.a

nnua

lrev

iew

s.or

gby

Uni

vers

ity o

f C

alif

orni

a -

Sant

a C

ruz

on 0

9/16

/12.

For

per

sona

l use

onl

y.

Documents

Causal Mechanisms in the Social Sciences